1 problem

0 votes0 replies1 view
The sharp two-step convergence-rate conjecture for gradient descent-ascent
Let $F\in\mathcal{F}(L_x,L_y,L_{xy},\mu_x,\mu_y)$ , where $L_x,L_y,L_{xy}$ are the smoothness parameters, $\mu_x,\mu_y$ are the strong-convexity and strong-concavity parameters, and…
Conjectureopengradient-descent-ascent optimization saddle-point-problemsarXiv posted 2 days ago

The sharp two-step convergence-rate conjecture for gradient descent-ascent